DJ Patil on Using AI to Move Fast and Fix Things

MOLLY WOOD: Today we’re talking with DJ Patil. DJ isn’t just one of the foremost data scientists in the world. He’s literally the co-inventor of the term “data scientist.” After serving as project leader for the Threat Anticipation Project at the Department of Defense, Patil was tapped to be the United States’ first Chief Data Scientist in 2015. He oversaw data-driven initiatives like the Cancer Moonshot. Currently a general partner at Great Point Ventures, he has also held key roles at LinkedIn, Skype, eBay, and PayPal. We were excited to ask Patil for his insights on technological transformation in large organizations, AI’s potential to unleash the power of data, and how the technology can be brought to bear on the world’s thorniest problems. And now my conversation with DJ. Thanks so much for being here.  

DJ PATIL: Thanks. Glad to be here.  

MOLLY WOOD: This season, we’ve been hearing repeatedly that curiosity is becoming one of the most essential traits in the age of AI. You have been a strong advocate for fostering a mindset of curiosity, especially among leaders. Why do you think being curious matters so much right now? 

DJ PATIL: Curiosity to me is the most important thing. Now, people are talking about neuroplasticity, all these different terms that are out there. Fundamentally, I think the way we should think about it is curiosity, passion for learning, passion for understanding, trying things. And with AI, you’ve got this thing, it’s really, in many ways, it’s a chatbot. You’re just like, okay, what do I ask it? What it does is, its ability to go from what is big data, like all this information out there, to big knowledge. You can sort of ask it, you can get curious results out of it, and if you are really a curious person, you can get lost in there for a while, learning and finding creative understanding or ideas that are there. 

MOLLY WOOD: So, you have steered ambitious technology organizations in government and enterprise and startups. What are the common pitfalls when it comes to adoption and implementation? I mean, I suspect that curiosity is going to be a part of that too, right? But that’s the key, is how, what do you do when you’re faced with that blank screen?  

DJ PATIL: The greatest lesson I’ve ever learned starts with listening. Curiosity isn’t about pontificating. It’s not about just trying to go out there and do stuff. It’s about learning. It’s about listening. Every job that I’ve been thrown into—it’s very rare that I’ve gotten myself into a job where I actually understand what I’m doing, to be very frank. How do you actually help people with government? It starts by listening to people and trying to understand what their needs are. And then you ask, what can technology do for them? And it’s kind of two directions that you come from, which is the raw science, the raw intellectual research kind of innovations, but then you have to ask, well, what does the instantiation of it need to look like to actually help people? Can I share a story actually about this from the White House?  

MOLLY WOOD: Always, always share stories from the White House.  

DJ PATIL: So I’ll tell you, one of the times I got to see the president really, really unhappy with me was when we went into the Oval Office to talk about the Precision Medicine Initiative. It’s the largest database of genomic information, healthcare information, all of this stuff. And so one of the things he was very emphatic on was, you have to do this with the people. I said, well, sir, met with this group that represents this rare disease, represents this group, this cancer, this type of thing. He’s like, I thought I was clear. You can’t do this through proxies of groups or other things. And so, you know, he asked like, what have you done to make sure we know what different communities are thinking about? And then that’s the moment you know your day is not going to go well, because everyone else in the room, in the Oval, everyone else is like just sitting there, very quiet, very stoic. But they’re all thinking like, ooh... 

MOLLY WOOD: DJ’s in trouble... [Laughter] 

DJ PATIL: Like, let’s see where this goes. And he pointed out, he’s like, I thought I was clear. You have to sit down with the people. And so we went out on the road and we met people in communities that we weren’t members of. And I remember being in Pittsburgh and we were in this town meeting with a bunch of people from different parts of the community. And this one elderly woman was in the back corner and she just said, do you really want to hear from me? And she just lit us up. She’s like, have you thought about this, this, this, this? And I remember walking out of the meeting, that was tough. And I was like, can we hire her? Like, she clearly understands a problem in a way we don’t, and we’re never going to understand her completely. We need to just find a way to build with her. We can’t build for her. We have to build with her on these things. And that’s one of the seminal challenges that I think we’ve had. You have to really build it with people to understand how to actually solve concrete problems. People ask, how do we use AI better? And if I had to tell people one thing, go sit with your people you’re going to be impacting. That could be an internal tool, could be an external service, could be, you know, something else. One of the framings that I’ve found really helpful that I like to use is, I like to say, technology is neither radical nor revolutionary unless it benefits every single person. 

MOLLY WOOD: Let’s put that anecdote actually in the language of data. Like, you’re a company, right? You have data, you have data scientists. You maybe have silos in your data. You maybe have a mess of it because you’ve just collected everything in the hopes that AI was going to come along. What is the role of the data scientist in this moment? How has it changed?   

DJ PATIL: So I’ll tell you the first thing that has not changed. You’ve got this data that’s broken up in silos all over the place, unstructured, messy, it’s a disaster. And then somebody says, make AI happen. And you’re just like, ta-da, and everyone’s like, well, the AI will just fix it. And you’re seeing this happen at scale where people are just trying to deploy AI and say magic will happen. It doesn’t work that way. The amount of time and energy that goes into building the foundation to allow you to use data effectively, the data scientist is one of the core advocates for what needs to happen there. Data scientists spend 80 percent of their time cleaning data. Doing stupid, boring problems—yucky stuff. What we’re still seeing is all that time and energy doesn’t actually go into being able to do the higher-order functions—the insights, the curiosity. The way the job is changing, and some of the things that I’m really excited about, is you get to use this partner of this AI system, this large language model, to say, well, what are other interesting questions I might want to ask? Then there’s this other side of this, which is to know when to ignore the data. You know, you can be so data-driven that your data’s stupid. That’s a real issue. And so the way to do that is create a hypothesis. You create that hypothesis and then you try to figure out how do you make it testable as a business. And one of the things that’s happened, there’s shocks all the time to businesses, economic situations happen. Natural disasters happen. That’s in a natural experiment. The question is, what do you learn from it? And if you’ve instrumented the data and you’ve got the data in the right place, now you can find interesting insights because of that. And then part of that is then really spending time in that community to get context. And the best data science stories are ones always where the data scientist is tethered tightly with domain experts or is in the field. This could be data scientists working in a city to help the fire department. And this is the same thing that I try to do in the companies that I work with, is embed the technologists, embed the data scientists directly with the people who are on the proverbial front lines, so that then we make it a we problem, not an I problem. 

MOLLY WOOD: Much like you do not want to silo your data, you don’t want to create an AI team. You don’t want to create a pilot that no one sees except for a tiny group of people in an organization.  

DJ PATIL: That’s right. The biggest mistake that we see is people treat data and data science as a throw-it-over-the-fence approach. You’re like, look, we got a question. Throw it in there, just like it was some—it’s like, instead of typing it into the LLM, you kind of hand it to some data scientist to go work on. That doesn’t work. You have to be deeply embedded and curious with them to—like at LinkedIn, one of the cultural tenets that we created, and I’ve done this ever since, is we just created a data meeting. No decisions were ever made. You’re there to just ask questions and just try to understand. Somebody might say, well, I see the data this way. Somebody else says, I see the data this way. If you have a decision at that, politics come in. Here, you have the ability to now ask and sort of just get smarter. The goal is to level everybody up. That’s the fundamental goal. And there’s a version of this that happens in the White House. They aren’t decisional meetings. They’re informative information meetings, and there’s an informational memo that goes out. It’s not a decisional. And because the people who need to do the decision—they may be coming from other places, they may be coming from other parts of government or the business. And so there’s a different meeting for that. 

MOLLY WOOD: I mean, I don’t understand how I’ve spent, uh, an undisclosed number of years on this planet and never worked at an organization that has meetings just for information as opposed to decision, and delineates them that way. Like, that all by itself feels like a huge unlock. And then go back to your thought partner, whoever it is, the decision maker or your LLM and say like, okay, here’s all the data that I have, what are some possible decision trees?  

DJ PATIL: That’s right. And the number of times where I have personally been in a meeting where, as the data person, have advocated, we should do something that the data doesn’t suggest, is because we’re human learning machines. Then we process that data and we turn it into output. And so we have this idea and frames, gestalts in our mind. What we then have to do is, how do we take new data and update our priors or our way of thinking about it? And that’s an experimental, but because of that, sometimes we have to say, look, the strategic thing can’t be captured in data. But it’s the right thing to do. Maybe it’s a moral thing, maybe it’s an ethical thing, all these other things. But then, and this is why like one of my mentors, the late Secretary of Defense Ash Carter, we had the first memo on this for autonomous weapon systems and AI, [and it] w,as humans have to be in the loop because judgment has to happen. The information may say, do this, but you as a human say, you know what? That’s a catastrophic line that we should not cross. We should not do this. And, you know, we’ve made famous movies about this of, like, where somebody’s single judgment is what saves the day.  

MOLLY WOOD: So now you are a partner at a venture capital firm. You live in the future in many ways, right? Because you see the newest and best thinking, and I would imagine you are seeing companies that are what we like to call AI-native, that are, you know, creating and forming themselves and thinking about their futures in totally different ways. Are there through-lines that you’re starting to see?  

DJ PATIL: Yeah, so this is why I call things actually AI native is because, you know, I grew up in a world pre-desktop native, pre-laptop native. And so I would call myself even mobile fluent, you know, like, because the people who are truly mobile native were the people who grew up with phones in college. And so now we’re in a moment where we have this interesting group of population that are really in their freshman, sophomore year of college, I would argue. They’re the ones who are going to be AI native. So AI native is coming. It’s not even here yet. The rest of us are going to be AI fluent. What is starting to show up is a question of, what does an AI-native company look like? What does an AI-fluent company look like? And what is something that just doesn’t understand that it doesn’t need to exist or it doesn’t have the ability to pivot into what it looks like in an AI world? And I would make this a broader statement that is not just about companies, this is about society as well. We think about society like recessions. Do we have a soft landing? Do we have a hard landing? As humans, we’re really bad at two things: We’re bad at exponentials, as we saw with COVID and other things, we’re really bad at understanding exponentials. And we’re really bad at transitions. Well, AI is on an exponential trajectory, and it’s on a rapid transformational arc. So are we going to have a hard landing or a soft landing for society, companies, and all those things? And there’s one hat, which is as an investor, you’re looking at it as, what are the next things that are going to be the disruptive replacements and opportunities to make the world happen? There’s another portion which says, how do we also make sure that there are technologies, companies that aid in this transition and have an opportunity to do things.

MOLLY WOOD: We’re talking about AI-native companies now, but we don’t even have full implementation of agentic AI. There’s yet another revolution potentially on the table, and I wonder how you think about the adoption of that, the integration of that. Can you even call yourself AI native if AI itself is going to potentially dramatically change in another month?  

DJ PATIL: Maybe this is the one way I think about it right now. Last calendar year was really spending pennies on AI. If you look at enterprise adoption, true enterprise adoption, it’s still very early. So last year we went from paying pennies; this year we’re kind of shifting to saving pennies. Maybe we’ll get to nickels by mid this year, maybe towards the end of the year, and really hopeful it’ll be in quarters. 2026, 2027 is when we shift to seeing potentially dollars on the balance sheet for those that do it well. And so that’s the way I think of this in the arc. The technology will transform radically. The one way to think of this is, in a typical year for a company, you might get one big product release. You know, they have the big announcement, there’s like one big thing. In startups, maybe you’ll get three product releases in a year. In AI, we’re at 10 iterations of technology releases in a year that are shifts. Could be model size, could be applications, could be stuff like, how does it work with agentic, all those things. We’re on an exponential arc of how to think about the iteration of that tech. And the cultural transformation is operating at a 1x speed. And so the place where I tell people, if you want the leverage, what is going to give you the leverage? It’s not technology. It’s culture, it’s people. It’s helping people know how to use this technology to be more adept. And I got to see this firsthand, at some of the slowest places ever—I was working in national security when Colin Powell was at the State Department, and he gave a memo that said everyone should have a desktop computer. That sounds like crazy now, right? But everyone’s like, oh, the costs, the costs are going to be insane. And like, what are we going to, can we trust people with these computers? And what happens if they take—you know, there was all these questions, and then now you think about it, you’re like, how did you do your work? That was a cultural transformative moment. Those are the versions that we’re going to need to see. It’s like, what does it mean to train people? What does it mean to give access to these technologies? How do we put the safeguards in them to help people make sure they don’t do something that’s going to be potentially harmful to themselves or to the company or the organization. 

MOLLY WOOD: On that note, I think, I mean, there’s a tendency, as you are well aware, in tech to move fast and break things. I believe you have an alternate phrase.  

DJ PATIL: We do, it’s from a book that we wrote on data and ethics, AI and ethics. This adage needs to be, move quickly and fix things. And what I think is, and why it resonates with people, is there are types of people out there who run into burning fires—firefighters, first responders, all these. There’s a form of technologist out there that’s a first responder, and they’re eager and excited to actually figure out how to use their skills to help improve situations for everybody. And that doesn’t mean it has to just be a nonprofit or something like that. It’s, how do we be intentional? How do we be thoughtful about what we’re trying to actually do, and actually make sure what we’re building is really what our intention is. I don’t want to make it seem like, you know, someone would say, oh, DJ’s brakes, like, we gotta go all gas, gas, gas, gas versus brakes. I’m very much about gas, but I think it’s a question of like, how can we do it? Because I’ve spent a lot of time with parents of children with rare diseases and kids with rare diseases, and I know there is no time. We cannot be wasting any time for those families with data silos, some of the data protections that are incredibly thoughtful, well-intentioned, but it doesn’t matter for that kid whose life expectancy is so short and could be extended, if not saved. And so, where do we find the right balance of that is being intentional about what problem we’re trying to solve.   

MOLLY WOOD:  I tell my son all the time, actually about driving, that there is a way to move with haste and not carelessness, which I think is part of what you’re saying. And related to that, we are in a moment of extreme uncertainty. We do not know what’s going to happen. We have these tools that some people say are contributing to that uncertainty. Other people say these tools can help us navigate through it. What do you think is the role of technology in this moment?  

DJ PATIL: I am incredibly excited about these technologies from the perspective of how they might be able to help people. I think, for example, one of the things is people have deep curiosity when they have an issue and they’re trying to figure out, you know, something, it might be a medical issue, it might be a social issue, like where’s our peer groups, where are our other systems? And a lot of these systems have atrophied. We need to bring those up, but we also can use technology to support on these fronts. I’ve met plenty of people with terminal diseases who are using LLMs to figure out how to advance their care, because when they talk to the physician or the care team, they’re getting insufficient answers. That’s an easy one that we hear about a lot of times. But I also hear about the people who, you know, are just afraid and anxious going into a job interview. One that I just saw the other day firsthand was my son. He is, you know, doing his interviews for the very first time and he is like, what do I do? And I was like, well, let’s just go construct a set of interview questions. This is something we might buy a book for before. Might be something we might have, you know, done a web search for, or something else. But here we’re able to really go, ah, here’s a set of questions. And then he wrote down his answers and we critiqued it. And so in a short span of time, he got really good. And then the follow-on, and this is the key thing, was don’t just settle there, now do this with a real human. Go talk to your roommate, have your roommate ask you them, see what that response is, see how it marries up, and then flip the script and go through it vice versa. I think that’s where we then use this technology to augment us, rather than just sort of as an intervention or a crutch.  

MOLLY WOOD: You introduced a kind of back-of-the-napkin framework it sounds like at the White House that went a little bit viral. Explain that to us.   

DJ PATIL: The backstory. Anybody can, if you want the longer version, you can see it on my LinkedIn page of the history of where this came from. But it came during a very tense moment after we’d seen a number of people of color, predominantly Black, who’d been killed in tragic law enforcement situations, along with officers who’d been really hurt in, you know, escalations and other environments. And so we brought people together. It was, what could we do with technology to help in these situations? What could that look like? And so these were a set of principles that I’ve used in building things that are there. And so one of them is you start by prototyping for 1x, building for 10x, and then engineering for 100x. And those X’s are scale. So start with prototyping. Don’t go to the 100x solution right away. Start with something really small, very tangible. See what you can do, and then make it bigger over time. Another one is, you know, how do you plan and work and execute? And so you plan, really the goal is at this—the pithy version of it, to sum it up is ship daily, ship something, everything daily. But you want to plan in years, and planning in years but shipping in days is the key thing. And the final one is, what does it take to double the impact while cutting the timeline in half? And this, for the math nerds out there, they’ve already got it. This is a hack because it no longer puts you on a linear paradigm. It puts you on a log-base-two paradigm, so it starts forcing prioritization exponentially. So only the most important things happen in that way. And when you start thinking about it that way, that’s how you get to solving problems at scale. And that I find is one of the most impactful ways to take something that you’re working small and just make it really big and the best answer for that is, it’s a classic African proverb. If you want to go fast, go alone; if you want to go far, go together.  

MOLLY WOOD: If your listeners could take away one actionable AI-related insight from you, what would you want it to be?  

DJ PATIL: Don’t just live in the AI system. Get out of the AI system and get into the real world. And then use that, use the AI system, to help bridge you into the real world and figure out what it’s like. But don’t just be stuck in front of the screen. Get out there, get with real people, talk to them, understand what’s really going on, and use the AI to help augment that.  

MOLLY WOOD: And then what are some surprising ways that you maybe are using some of these tools at work or even in your personal life? 

DJ PATIL: If anybody follows me on social media, they know I do a lot of photography, and so one of the things that I do is, actually, I upload my images to the LLM and I ask it to critique it, but I don’t just ask it to critique it. I ask it, say, your judges are these different types of photographers who are historical. I ask them to judge it, and then I ask them to give me suggestions on what I could have done differently on composition or post-processing to improve it. And so it’s kind of acting as this coach. I then take that and I actually do show it to real humans and ask them, like say, okay, this is what I got from the system. What do you think? And they always have some, they’re always like, they give me some additional insight that can’t be fully captured. But I found it to be a really effective, helpful coach in this sort of strange way, and this is like one of those things I was like, I was playing one day and I was just like, I wonder if. And I just did it and I was like, whoa, that was weird. That’s, I didn’t expect that. And so a lot of, I try to spend, carve out a little bit of time for what I would call play and just playing with these things and just trying to do that. And I just want to emphasize, I know that is a statement of privilege that not everyone else has, but finding ways where you can do that a little bit on the edge is something that I would advocate for everyone. 

MOLLY WOOD: Fast-forward three to five years, what do you think will be the most profound change in the way we work?  

DJ PATIL: We won’t think about AI. We don’t think about, like, what does the mobile team do if you’re inside a company, like, everyone’s doing mobile. We don’t ask like, oh, when we go, you know, we try to do something or interact with something digitally, we don’t go, oh, did you go to the web version on your desktop? Or did you go do the mobile version? It’s just integrated. And so I think a lot of these interfaces, a lot of the products that will show up over the next five years will be ones that will be so naturally embedded that we don’t think of it as the AI system versus the problem that we’re trying to solve. It kind of sits side by side.  

MOLLY WOOD: Perfect place to end. DJ Patil, America’s first chief data scientist, and now a general partner at Great Point Ventures. Thank you so much for the conversation.  

DJ PATIL: Thanks. Thanks for having me. 

MOLLY WOOD: Thank you all for joining us, and keep checking your feeds. We have more fascinating guests on the way with actionable insights that can help leaders develop an AI-first mindset, leverage data to make smarter decisions, and maximize the ROI of AI. If you’ve got a question or a comment, please drop us an email at [email protected], and check out Microsoft’s Work Trend Indexes and the WorkLab digital publication, where you’ll find all our episodes along with thoughtful stories that explore how business leaders are thriving in today’s new world of work. You can find all of that at microsoft.com/worklab. As for this podcast, please, if you don’t mind, rate us, review us, and follow us wherever you listen. It helps us out a ton. The WorkLab podcast is a place for experts to share their insights and opinions. As students of the future of work, Microsoft values inputs from a diverse set of voices. That said, the opinions and findings of our guests are their own, and they may not necessarily reflect Microsoft’s own research or positions. WorkLab is produced by Microsoft with Godfrey Dadich Partners and Reasonable Volume. I’m your host, Molly Wood. Sharon Kallander and Matthew Duncan produced this podcast. Jessica Voelker is the WorkLab editor.